ABSTRACT
This research work is aimed at modeling the causes of death among adult patients aged 15 years and above admitted at Federal Medical Centre (FMC), Owerri, Imo state. The data used for the study were extracted from record available at the Data Unit of Federal Medical Centre, Owerri, Imo State, and covered the period, 2013 – 2018. The data were analyzed using descriptive and multinomial logistic regression methods. 368 out of the 998 patients admitted died, implying a mortality rate of 369 per thousand of such patients. The average mortality rate was higher in the females (372 per 1,000 patients) group compared to male,(366 per 1,000 patients) group. The sex ratio shows excess female deaths over male deaths in the two youngest age groups 15-29 and 30-44, while the reverse was the case in the older age groups. Four major groups of diseases accounted for 70 percent of all deaths among the patients; they include Cardiovascular diseases, Infections diseases, Genitourinary diseases, and Digestive disorder. The most fatal of these groups of diseases is Genitourinary diseases with a fatality rate of 527 deaths per one thousand cases, which seems to be higher for males (578 deaths per 1,000 cases) than for females (407 deaths per 1,000 cases). The result of the multinomial logistic Regression revealed that Genitourinary group of diseases has the highest probability (0.7672) of causing death among adult patients admitted in Federal Medical Centre, Owerri. Other less fatal diseases include Malignant neoplasms with a probability of 0.0802, and Digestive disorder with probability of 0.0522.
TABLE OF CONTENTS
Title Page i
Declaration ii
Certification iii
Dedication iv
Acknowledgements v
Acronyms vii
Table of
Contents ix
List of
Tables x
List of Figures xii
Abstract xiii
CHAPTER 1: INTRODUCTION 1
1.1 Background of the Study 1
1.2 Statement of Problem 5
1.3 Aim and Objective 5
1.3.1 General Objective 6
1.3.2 The Specific Objective 6
1.4 Significant of Study 6
1.5 Scope of Study 7
CHAPTER 2: LITERATURE
REVIEW 8
2.1 Conceptual Framework 8
2.1.1 Leading cause of death 8
2.1.2 Causes of the sudden natural death in
various hospital 11
2.1.3 Maternal mortality in Nigeria hospital 11
2.2 Empirical Review 13
CHAPTER 3: METHODOLOGY 18
3.1 Sources of Data 18
3.2 Variable Specification 18
3.2.1 The response variable and baseline category 18
3.2.2 Explanatory variables 19
3.2.3 Model building 19
3.3 Method of Analysis 20
3.3.1 Age- sex structure of the research data. 20
3.4 Multinomial Logistic Regression Model 21
3.4.1 Estimating response probability of cause of
death among patients aged 22
years
and above .
3.4.2 Parameter estimation 25
3.5 Assumptions of Multinomial Logistic
Regression 28
3.5.1 Goodness-of-fit test 28
3.5.2 Test for multicollinearity 30
3.5.3 Variance inflation factor (VIF): 30
3.5.4 Remedial measures for multicollinearity 31
3.5.5 Classification table of the polytomous
predicted variable 31
3.6 Estimating the Wald test statistic 33
3.7 Estimating the Pseudo R2 test
statistic 34
CHAPTER 4: RESULTS AND DISCUSSION
4.1 The response variable categories 36
4.2
Determining the Leading Causes of
Death In FMC Owerri 39
4.3: Result
for Goodness-of-Fit-Test 46
4.4 Tests
for Multicollinearity 47
4.4.1 Examination of the Pearson correlation 47
4.4.2 Multicollinearity test using variance
inflation factor (VIF) 48
4.5 Multinomial
Results 49
4.6
Predicting The Probability of Dying
of Each Cause of Death 60
4.7 Fitting
A Statistical Model to Causes of Death Data Among And
Patient Aged 15 Years and Above 63
4.7.1 The estimated result of the model fitting
information 65
4.7.2
Likelihood Ratio Tests for selected model 66
4.8 Results for Pseudo R-Square 68
CHAPTER 5: CONCLUSION AND RECOMMENDATIONS
5.1
Conclusion 69
5.2
Recommendations 70
References 71
Appendix 76
LIST OF TABLES
3.2: Explanatory variables and their categories
19
4.1: Frequencies of the response variable
categories 36
4.2: Distribution of patients aged 15 years and
above admitted in
FMC, Owerri and deaths among them, 2013-2018. 38
4.3: Levels
and trend of mortality rates among patients aged 15 years
and above admitted at FMC, Owerri
2013-2018. 39
4.4: Distribution of death by cause, place of
residence, for the period
2013-2018,
Federal Medical Centre, Owerri. 40
4.5: Cause of death structure for patients aged 15 years and above
admitted in FMC, Owerri 2013 -2018. 41
4.6: Distribution
of deaths by age and sex period 2013-2018
at
FMC, Owerri. 42
4.7: Fatality
rate of diseases for which patients Aged 15 years and
above were admitted in FMC, Owerri 2013-2018 44
4.8: The estimation of the deviance and Pearson’s
chi-square
goodness-of-fit
test 46
4.9: Pearson
correlation for the explanatory variables 47
4.10: Estimated collinearity statistics (tolerance
and VIF) of explanatory
variables
for patient age 15 years and above, who were admitted at
FMC, Owerri
2013-2018. 48
4.11: Logit
coefficients from multinomial 50
4.12: Logit coefficients, of multinomial
logistic regression of dying
of
1of 9 causes Vs dying of other causes, on selected predictor
among
patients aged 15 years and above admitted in FMC,
Owerri, 2013-2018. 57
4.13: Estimated
probability of dying for each response variable
category among
patients Age 15 years and above admitted in Federal
Medical
Centre Owerri, from 2013-2018. 63
4.14:
Estimated step summary of fitted model 65
4.15:
Estimated model fitting
information 66
4.16: Estimated likelihood ratio tests for selected
model of the patient
age 15years and admitted in the Federal Medical
Centre Owerri,
2013-2018. 67.
4.17:
Pseudo R-Square 68
LIST OF FIGURES
4.1: Percentage distribution of total deaths by cause according to
sex
among patients age 15 years
and above, FMC, Owerri: 2013-2018. 42
4.2: Case fatality rate by sex among patients
aged 15 years and above, FMC,
Owerri: 2013-2018. 45
ACRONYMS
WHO: World Health Organization
ICD: International Classification of Diseases
MMR: Maternal Mortality Rate
SRD: Sex Ratio at Death
MNLR: Multinomial logistic Regression
FMC: Federal Medical Centre
CVDs: Cardiovascular Diseases
VIF: Variance Inflation Factor
AIC: Akaike’s Information Criterion
BIC: Bayesian Information Criterion
CFR: Case Fatality Rate
MR: Mortality Rate
Ca: Cancer
CCF: Congestive Cardiac Failure
HHD: Hypertensive Heart Disease
TB: Tuberculosis
STDs: Sexually Transmitted Disease
HIV: Human Immunodeficiency Virus
CKDD: Chronic kidney Disease
KIDD: Kidney Disease
BOO: Bladder Outlet Obstruction
UTIs: Urinary Tract Infection
CLD: Chronic Liver Disease
GOO: Gastric Outlet Obstruction
Obst.Lab.: Obstructive Labor
DFU: Diabetes Foot Ulcer
PUD: Peptic Ulcer Disease
Dm: Diabetes Mellitus
Leuk.Ca: Leukemia Cancer
CPR: Cardiopulmonary Resuscitation
CHAPTER
1
INTRODUCTION
1.1 BACKGROUND OF THE STUDY
Human death has always been shrouded by
mystery. In modern times, however, the study of death has become a central
concern. According to Siegel and Swanson (2004), death is the permanent
disappearance of all evidence of life at any time after live birth has taken
place. The concept of death is key to understanding of the phenomenon (Mohammad
and Gilblert, 2010). Determining when death has occurred is difficult, as
cessation of life functions is often not simultaneous across organ systems
(Henig, 2016).
Historically, attempts to define the exact
moment of a human’s death have been subjective or imprecise. Death was once
defined as the cessation of heartbeat (cardiac arrest) and of breathing,
but the development of cardiopulmonary resuscitation (CPR) and prompt
defibrillation have rendered that definition inadequate because breathing and
heartbeat can sometime be restarted.
The death
of a person has legal consequences that may vary between different
jurisdictions. A death certificate is issued in most jurisdictions, either
by a doctor, or by an administrative office upon presentation of a doctor's
declaration of death. For us to ascertain the cause of death of a person, we make use of death
certificate, which provides information on medical condition that led to death.
The information is coded using standard cause of death classification
categories developed by World Health Organization (Rasika et al., 2014).
Reliable
information on deaths by cause is a vital ingredient for planning, managing and
monitoring the performance of the health sector of any nation. Estimates of
mortality rate disaggregated by age and sex for specific causes provide insight on the evolution of the overall
mortality rate in a population (Murray and Lapez, 1997).
In
most countries, death certificates constitute the largest disease related
source of information for public health research and policy making. Many middle
income countries, including Nigeria have established vital registration systems
to compile their mortality statistics.
Leading
causes of death are underlying causes of deaths that usually account for large
number of deaths within a specified geographical area and time period. Furthermore, the leading cause of death
statistics help health authorities determine the focus of their public health
intervention. For instance, a city or country in which deaths from heart
disease and diabetes rise rapidly over a period of few years has a strong
interest in starting a vigorous programme to encourage healthy lifestyles to
help prevent these illnesses. Similarly, if a city recognizes that many
children are dying of pneumonia, but only a small portion of the budget is
dedicated to providing effective treatment, it can increase the spending in
this area.
In the 1960s and 1970s, it was common for observers to
speculate about a global convergence in mortality patterns (Omran 1971a;
Stolnitz 1965; UN 1975). The optimism was based on the diffusion of medical
knowledge and technologies in the post World War II-period, which facilitated
faster improvements in the life expectancy of developing countries compared to
the eighteenth and nineteenth century mortality transitions in Western Europe
(Davis 1956; Omran 1971b). Over the last two decades, about one out of four
countries in the world experienced a mortality crisis and decreasing life
expectancy due to conflict (e.g., Rwanda, Angola, Sierra Leone, Liberia, Iraq,
Somalia), economic crises and the failure of health systems (e.g., Russia,
Kazakhstan, Belarus, Ukraine, Democratic People's Republic of Korea, Zimbabwe),
and, most importantly, because of the mortality impact of the HIV/AIDS epidemic
(UN 2009). The paucity of vital statistics is problematic for estimating adult
mortality while information on infant and child mortality is in principle easily
elicited from the mothers. In some countries, adult health has drastically
deteriorated since the 1980s, leading to an increasing heterogeneity in adult
mortality levels. Information on causes of death suggests, however, that the
extremely high adult mortality levels in some of the South Eastern African
countries are not the sole result of the HIV/AIDS epidemic, but due to the
triple burden of infectious and chronic diseases (Reniers et al. 2014). A review of adult mortality trends in Africa inevitably
induces controversies about data sources and methods of estimation, because of
a lack of reliable data for estimating adult mortality. The crux of all
difficulties in estimating adult mortality in most African countries is the
absence of an accurate vital registration system. Apart from the northern
African countries, only Mauritius, Cape Verde, Réunion, and South Africa have
consistently provided nationally-representative vital statistics over the last
few years (Cleland 1996; Mahapatra et al.
2007; Mathers et al. 2005; Setel et al. 2007).
The leading causes of death in developed countries are atherosclerosis (heart disease and stroke), cancer, infectious disease and other diseases related to obesity and
aging. By an
extremely wide margin, the largest unifying cause of death in the developed
world is biological aging (Aubrey and Grey, 2007) leading to various
complications known as aging-associated
diseases. These conditions cause loss of homeostasis,
leading to cardiac
arrest, causing loss of oxygen and
nutrient supply, causing irreversible deterioration of the brain and other tissues. Of the roughly 150,000
people who die each day across the globe, about two thirds die of age-related
causes. In industrialized nations, the proportion is much higher, approaching 90%
(Aubrey and Grey, 2007). With improved medical capability, death has
become a condition to be managed if the
leading causes are known (Aubrey and Grey,
2007). Moreover, leading causes of death can be determined even when
reliable population estimates are lacking, making it a readily available
measure for health, wherever effective mortality reporting is in place. There
are a number of measures that can be used to gauge the relative importance of a
specific cause of death. These include age-adjusted death rates, cause
–eliminated life tables and cause –associated years of productive life lost.
Finally,
Anderso and Smith (2005) on behalf of World Health Organization developed a revised
version of cause of death grouping for determining underlying causes of death
for the last International Classification of Diseases (ICD). It is important to
obtain and follow the cause of death groupings that match the ICD revision
reflected in historical mortality data. This ensures comparability of
cause-of-death statistics within and between nations. They are also organized
to be able to compare leading causes across the ICD revisions used
historically.
The logistic model
is a statistical model that estimates the likelihood of an event occurring as a
result of the interaction of one or more independent variables. There are
different types of logistic model which
include : firstly Binary Logistic Regression: This type of logistic regression
is used when the dependent variable has only two possible outcomes, such as
"yes" or "no", "true" or "false", or
"success" or "failure". Binary logistic regression is
commonly used in medical research, social sciences, and marketing to model
binary outcomes such as disease diagnosis, voting behavior, and customer
churn(Hosmer, et al.,2013). Secondly, Ordinal Logistic Regression is the type of logistic regression used when the
dependent variable has three or more ordered categories, such as
"low", "medium", and "high", or "poor",
"fair", "good", and "excellent". Ordinal logistic
regression is commonly used in fields such as psychology, social sciences, and
education to model outcomes such as academic achievement, job satisfaction, and
quality of life (Long, 1997). Thirdly, Multinomial Logistic Regression is used
when the dependent variable has three or more unordered categories. Multinomial
logistic regression is commonly used in fields such as public health,
education, and psychology to model outcomes such as educational attainment,
disease severity, and mental health status (Agresti, 2019). Multinomial logistic regression is a
statistical technique that is used to model outcomes with more than two
categories. In this study, the outcome variable is the cause of death, which is
classified into nine categories based on the International Classification of
Diseases (ICD-10) codes.
1.2 STATEMENT
OF THE PROBLEM
For
long, adult mortality remained a neglected public issue in Africa (Bradshaw and
Timaeu, 2006). Previous studies focused on causes of death due to sudden
natural deaths (Obiorah and Amakiri, 2012), maternal mortality (Olopade and
Lawoyin, 2008) except Dudley and Hosik,
(2008) who carried out a study on
leading causes of death among elderly people in US. Recently, it has been shown
that the leading cause of death varies from region to region (Myunggu et al. 2020).It is important to
note that most deaths are preventable if the underlying causes are known.
Adequate planning to reduce mortality requires understanding the leading of
causes of death which help in optimum use of limited resource to reduce the
high death rate.
The
problem is to fit a statistical model that can accurately predict the cause of
death based on set of predictor variables. Specifically, the goal is to use a
multinomial logistic regression approach to analyze large dataset of deaths and
their associated characteristics, such as age, gender, marital status, medical
history, and other factors that may be relevant to understand the causes of
death.
The
model must be able to classify each death into one several possible categories
of cause of death (e.g., heart disease, cancer, accidents, etc), bases on the
available data. The challenge is to identify the most important predictor
variables to fit a model that is both estimable and interpretable, allowing
researches to better understand the underlying factors contributing to each
death.
Since
the outcome variable of interest is multiclass (i.e. has more than two levels)
and categorical (i.e the variables have no natural ordering), a multinomial
logistic regression technique is considered appropriate. This is because
multinomial logistic regression provides highly interpretable coefficients that
quantify the relationship between the independent and the outcome variables. In
addition, multinomial logistic regression is more flexible than ordinal
logistic regression because it dose not require many strong assumption about
the structure of the data (Kwak and Clayton-Matthew, 2002).
It
is for the above reason that this study is conceived to identify and model
using multinomial logistic regression the major causes of death among patients
aged 15 years and above who were admitted to the Federal Medical Centre,
Owerri, Imo State.
1.3 OBJECTIVES
OF THE STUDY
1.3.1 General
objective:
This
study is aimed at identifying and modeling the leading causes of death among
patients aged 15years and above admitted to Federal Medical Centre, Owerri, Imo
State.
1.3.2 The
specific objectives include:
a) To
identify the causes of death structure among patients aged 15 years and above
admitted at Federal Medical Centre, Owerri, Imo State.
b) To
fit a multinomial logistic regression model to causes of death data among
patients’ aged 15 years and above.
c) To
predict the probability of dying by each cause of death identified among
patients’ aged 15 years and above.
1.4
SIGNIFICANCE OF STUDY
This study
provides information on the major health challenge affecting the population
aged 15year and above in the Federal Medical Center Owerri, Nigeria, using a
multinomial logistic regression approach. Nigeria, like many sub-Saharan
African countries, has a high mortality rate, and the Federal Medical Center
Owerri is one of the largest tertiary hospitals in the country. Therefore,
understanding the causes of death in this hospital can provide insights into
the broader mortality trends in Nigeria and the region as a whole. It will be of utmost importance for
health policy formulation and decision making. Reliable
information on cause of death can guide research, optimum resource allocation
and help in effective management of health services, leading to improvement in
the health of the people and saving of lives.
1.5 SCOPE OF STUDY
This
study is limited to data on persons’ aged 15 years and above admitted to
Federal Medical Centre, Owerri, Imo State for the six years period of 2013-2018
and the use of Multinomial logistic Regression .
Click “DOWNLOAD NOW” below to get the complete Projects
FOR QUICK HELP CHAT WITH US NOW!
+(234) 0814 780 1594
Login To Comment